A Socioeconomic Edge Effect: Rational Crime in Small Areas of Sharp Socioeconomic Contrast

Abstract

Objectives:

To determine whether criminogenic “edges,” as defined by crime pattern theory, exist at points of sharp contrast of socioeconomic status (SES).

Methods:

The study uses a quasi-experimental design with pattern matching logic. A series of negative binomial regression models separately examine five different crimes with an economic incentive as dependent variables, and five crimes without an economic incentive as nonequivalent dependent variables, to determine whether census block groups of predominantly and comparatively higher SES than the wider surrounding area experience greater reported rational crime than would otherwise be expected.

Results:

The census block groups of comparatively higher SES located within and/or near areas of predominantly lower SES experienced one of the five crimes with an economic incentive, robberies by firearm, 40 percent more frequently than would otherwise be expected.

Conclusions:

The study’s findings are partially consistent with its hypothesis, which is grounded in crime pattern, rational choice, routine activities, and social disorganization theories. The findings encourage future research that may extend the definition of an “edge” under crime pattern theory as well as research at the intersection of criminological theories.

Keywords

crime pattern theory edge small area rational choice theory criminological theory social disorganization

The introduction of crime pattern theory by the Brantinghams in 1991 marked an important advancement in criminology; it was a concept they developed and refined further just a couple of years later with their discussions on the awareness and activity spaces that can be found in the nodes, paths, and edges of our everyday activities (Brantingham and Brantingham 1993b). Although nodes and paths are rather straightforward concepts, the definition of edges is more nebulous and imprecise, as it has been argued that, in addition to the obvious edges in the world around us, edges also exist where there is a change in land use (Song et al. 2017), residential density (Song, Spicer, and Brantingham 2013), or even territorial functioning and behavior settings (Taylor 1988, 1997). Research indicates that crime may be higher in edge areas (Brantingham and Brantingham 1993b; Song et al. 2017; Song et al. 2013). This study sets out to determine whether a criminogenic effect exists at socioeconomic edges, small areas where economic means significantly contrast with their broader surroundings. The study adds to the extant research in crime pattern theory and environmental criminology by examining the criminogenic effects that may exist in small areas (census block groups) of higher socioeconomic status (SES) that are situated within, or in close proximity to, much larger areas of predominantly and comparatively lower SES.

Theoretical Framework

The relationship between socioeconomic characteristics and neighborhood crime has been a subject of study going back at least to Shaw and McKay’s social disorganization theory in 1942. It has since been recognized that an area’s economic resources have an impact upon crime, as areas with fewer resources often experience greater crime as a result of a diminished ability to exercise informal social control (Bursik 1986, 1988; Kornhauser 1978; Sampson 2012). It has also been suggested that the criminological lens that examines what drives different crime rates at the community level might provide a clearer image (or, at least, a different perspective) when refocused at smaller levels (Andresen and Malleson 2011; Brantingham, Dyreson, and Brantingham 1976; Sherman, Garten, and Buerger 1989).

Environmental Criminology

Crime clusters spatially (Guerry 1833), and the nonrandom distribution of crime can be attributed to factors specific to the places where crime does (and does not) occur (Lersch 2007; Weisburd et al. 2016). The collections of theories that examine the spatial distribution of crime are generally grouped under the environmental criminology heading (Brantingham and Brantingham 1991).

The routine activities theory posits that when a person motivated to commit a certain crime (a likely offender) comes together with a suitable target in a certain place and time, absent a capable guardian, crime will occur (Cohen and Felson 1979). The rational choice perspective (Clarke and Cornish 1985) suggests that a person balances the potential costs of engaging in any particular behavior against the perceived expected benefits. This cost-benefit analysis takes the form of weighing the risk of being detected, arrested, and/or criminally charged against the reward of engaging in the specific criminal activity. In making what Cornish and Clark (1986) call the criminal event decision, when an offender opts to commit a specific crime at a specific time and place, upon a specific target, the decision is made using the totality of the offender’s knowledge given the benefits, the risks, and situational factors known to the offender at the time of the decision (p. 2).

Crime pattern theory focuses on a person’s routine activities and how they inform personal awareness and action spaces and, in turn, how these awareness and action spaces relate to that person’s decisions to commit crime where he or she does (Brantingham and Brantingham 1993a). A person’s normal, everyday activities typically take that person to the same places repeatedly, and he or she will frequently travel the same ways back, forth, and between. These nodes and pathways, as well as the immediate vicinities around them, will become part of a person’s awareness space as his or her knowledge about the areas grows (Brantingham and Brantingham 1993b). This personal knowledge will influence where that person may choose to commit a crime based upon his or her “perceptions of environmental cues that separate good criminal opportunities from bad criminal risks” (Brantingham and Brantingham 1991:3). Ultimately, offenders will develop a “template” through which to assign opportunity values to specific places and situations (Brantingham and Brantingham 1993a:263-64), and higher opportunity values will result in more crime at these places.

Edges

In addition to nodes and paths, in their discussion of places in which crime rates are at their highest, the Brantinghams (1993b:17) also highlighted what they termed edges: “places where there is enough distinctiveness from one part to another that the change is noticeable.” They provide a series of concrete examples, ranging from the edges of parks to rivers, to residential and commercial areas, that are plainly visible and readily apparent. They also discussed, however, citing the work of Taylor (1988), how edges can exist at the “spatial limits of mutual territorial functioning…or potential territorial conflict between different groups or land uses” (Brantingham and Brantingham 1993b:17) as well as at “the limits of perceptual comfort felt by outsiders entering unknown areas” (Brantingham and Brantingham 1995:12). These edges at the borders of mutual territorial functioning may not be clear or so readily apparent to outsiders but might perhaps be more apparent to insiders whose awareness and/or activity spaces provide them more intimate knowledge of the areas; regardless, they are no less existent merely because they are more difficult to see. Earlier work by the Brantinghams (1975:275) found that areas that marked transitions between a number of demographic and socioeconomic variables experienced increased reported residential burglaries, and more recent studies have also found that crime occurs more frequently in edge areas (Rengert, Lockwood, and Groff 2015; Song et al. 2017; Song et al. 2013).

In sum, a sharp and sudden increase in an area’s SES represents an edge. The area of this socioeconomic edge may be assigned a higher opportunity value by motivated offenders with awareness of the area, as some edge areas have been shown to experience greater reported crime. The higher SES of a socioeconomic edge offers potentially greater rewards for crime in which there exists an economic benefit to the offender, or “rational crime” (Covington and Taylor 1989:142), than the wider surrounding area of comparatively lower SES in which the edge is situated. At the same time, the crime-reducing mechanisms of informal social control and collective efficacy (Sampson 2012) that are often present in neighborhoods with high SES may not be present in these small edge areas, as they are situated in wider neighborhoods with predominantly much lower SES. Accordingly, any “edge effect” upon reported crime in these small edge areas would be separate and distinct from any effect of the overall level of SES. This study hypothesizes that the potentially increased economic rewards will draw rational offenders to small “socioeconomic edge” areas of higher SES nested within, or in close proximity to, areas of comparatively much lower SES, and this will result in increased reported rational crime in those small areas.

Methods

This study examined crime at points of sharp contrast (or “edges”) in terms of SES. The Brantinghams (1997), authors of crime pattern theory, have indicated that “for practical purposes individual criminal events must be aggregated in order to assess patterns” (p. 264), and in this analysis, the appropriate level of aggregation was the census block group level.¹ The city of Philadelphia offered an ideal study location, as it is a densely populated city comprised of 1,336 socioeconomically diverse census block groups, with a typical size of the census block groups such that if one contrasts with neighboring areas in terms of SES, a sizable portion (if not all) of that census block group may fit into the definition of an “edge” area as provided by the Brantinghams (1975, 1993b). The median area of the census block groups included in the analysis is 0.048 square miles, equivalent to a geographic square with side lengths of approximately 1,156 feet. Although there is currently no consensus as to a precise, quantifiable definition of what constitutes an edge, recent research suggests that edge effects may exist within 800 feet of the border of communities with contrasting sociodemographic characteristics (Rengert et al. 2015) or within 40 m (approximately 131 feet) of changes in land use (Song et al. 2017). This suggests that a substantial portion of each census block group (or in some cases the entire census block group) may in fact be characterized as an edge area, if such a contrast does in fact exist.

Factors specific to place impact the spatial distribution of crime (Weisburd et al. 2016), and these effects vary by crime type (Sherman et al. 1989). The criminogenic benefit in socioeconomic edges should appeal to rational offenders for crimes in which there exists an opportunity for economic profit; these edges may attract offenders of these crimes but would not provide any type of appeal, or perhaps even be apparent, to offenders of any other type of crime. Consequently, crimes were separated into two categories for study: “rational crimes” in which a rational offender balances risks against rewards in a decision-making process to achieve economic “gain” or benefit (Covington and Taylor 1989:142), and “nonrational crimes,” those committed for purposes other than economic benefit. The common reward in each rational crime is economic gain, but each crime has very different reward values and associated risks, which vary considerably by type of crime (Clarke and Cornish 1985). Crime pattern theory posits that offenders weigh criminal opportunities by simultaneously considering the totality of several factors pertaining to overall “target suitability” in light of each other (Brantingham and Brantingham 1993a:263), so as risks and rewards vary by crime type, the value of the socioeconomic edges will vary by crime type as well. As such, in order to test this study’s assertion that rational crime should increase in socioeconomic edges, but that nonrational crime should not, this study used a quasi-experimental design with a pattern matching logic. A series of different rational crimes, including an array of both violent and property crimes, were included as separate dependent variables to permit separate examination of the effect of the socioeconomic edge upon several different types of crimes. A selection of nonrational violent and property crimes was also included as nonequivalent dependent variables (Shadish and Cook 2009; Shadish, Cook, and Campbell 2002).

A concise explanation of the advantages of pattern matching research designs was provided by Shadish and Cook (2009), who indicated

they combine various design features so as to produce multiple probes of a causal hypothesis that inform different threats to internal validity without all sharing the same threat. When such probes all converge on the same effect, the plausibility of a causal inference is increased due to the more numerous alternative interpretations informed and the absence of shared bias. (P. 622)

Edge areas have been shown to experience greater reported crime generally, but this study contends that, due to the enhanced economic reward value socioeconomic edges present, these areas should experience increases only in rational crime, as they do not represent any enhanced reward for nonrational crime. The inclusion of nonequivalent dependent variables as part of a pattern matching logic improves upon this study’s internal validity by potentially demonstrating that the socioeconomic edges experience more rational crime than would otherwise be expected, but not other types of crime. The method offers an opportunity to reinforce the study’s theoretical contention that socioeconomic edges do represent enhanced rewards specifically for the economically incentivized rational crimes and that these rewards drive more rational offenders to these socioeconomic edges to commit rational crimes.

Data

ArcGIS Version 10.4 and Stata/IC Version 14.2 were used to manage and analyze the data, which were obtained from various sources.² The city of Philadelphia has 1,336 census block groups, and it was at this level that all data were aggregated and analyzed.³ Table 1 provides descriptive statistics for all variables in the analysis.

Table 1.

Descriptive Statistics for All Variables.

Variable	Mean	SD	Min.	Max.
Crime counts
Robberies by firearm	12.29	10.54	0.00	83.00
Robberies other than by firearm	16.49	18.89	0.00	282.00
Residential burglaries	33.81	20.08	2.00	196.00
Thefts from automobiles	52.82	56.74	3.00	857.00
Automobile thefts	15.33	12.58	0.00	144.00
Aggravated assaults by firearm	8.68	9.09	0.00	76.00
Aggravated assaults other than by firearm	22.24	17.87	0.00	155.00
Vandalisms	60.58	34.91	5.00	317.00
Arsons	1.81	2.37	0.00	28.00
Rapes	3.79	3.67	0.00	25.00
Spatially lagged crime counts
Robberies by firearm (lagged)	13.46	7.67	0.25	50.00
Robberies other than by firearm (lagged)	18.56	12.17	0.33	97.29
Residential burglaries (lagged)	34.68	13.57	7.20	113.33
Thefts from automobiles (lagged)	65.06	50.69	10.75	427.20
Automobile thefts (lagged)	17.30	10.30	2.00	72.33
Aggravated assaults by firearm (lagged)	9.22	7.09	0.17	42.25
Aggravated assaults other than by firearm (lagged)	24.29	13.34	2.00	90.50
Vandalisms (lagged)	65.79	23.85	18.00	165.00
Arsons (lagged)	2.07	1.67	0.00	11.80
Rapes (lagged)	4.18	2.78	0.00	16.00
Independent variables
High–low SES outlier (“SES-HL”)	0.02	0.15	0.00	1.00
Population	1,164.30	535.70	251.00	3,929.00
Percent White non-Hispanic	0.34	0.34	0.00	1.00
SES index	0.00	0.81	−2.49	2.13
Stability index	0.00	0.82	−3.67	2.35
Civic/institutional land uses	3.46	4.25	0.00	41.00
Commercial land uses	18.42	23.09	0.00	258.00
Residential land uses	362.23	169.65	0.00	1,585.00
Intersections with traffic lights	7.23	8.30	0.00	84.00
Intersections with stop signs	29.87	18.29	0.00	138.00
Subway/elevated stations	0.04	0.23	0.00	5.00
Major arterial miles	0.22	0.41	0.00	3.79
Minor arterial miles	0.34	0.48	0.00	6.73
Local street miles	1.09	0.75	0.00	7.81
Collector street miles	0.86	0.58	0.00	4.32

Note: N = 1,324. SES = socioeconomic status; SD = standard deviation; Min. = minimum; Max. = maximum.

Dependent variables

This study examined a total of 10 separate dependent variables (one dependent variable in each of 10 separate negative binomial regression analyses). Each of the dependent variables are counts of crimes that were reported to the city of Philadelphia Police Department during a five-year period, from January 1, 2010, through and including December 31, 2014, as recorded using Uniform Crime Reporting codes. The types of crime included as dependent variables were selected to permit an examination of a diverse sample of different violent and property crimes with different risks and rewards. Five of the ten dependent variables are counts of rational crimes (robberies by firearm, robberies other than by firearm, residential burglaries, thefts from automobiles, and automobile thefts), and the remaining five (nonequivalent) dependent variables are counts of nonrational crimes (aggravated assaults by firearm, aggravated assaults by other than by firearm, vandalisms, arsons, and rapes).

Independent variables

This study controlled for socioeconomic factors, the built physical environment, traffic and activity, population, and spatial autocorrelation. Census block groups that are statistically significant spatial outliers of high SES (relative to neighboring block groups) were identified and coded with a dummy variable (“SES-HL”), as these are the edge areas in which rational crime was predicted to increase.

SES index

Value is a consideration when offenders assess target suitability (Cohen and Felson 1979:591), and people with greater economic means are likely to have things of greater economic value. SES is an indicator of economic means (Braveman et al. 2005, Winters-Miner et al. 2015). This index variable (N = 1,324) is the average of five standardized (z-scored) variables (Cronbach’s α = .864): the natural log of the median household income of the census block group (N = 1,322), the natural log of the median average home value of owner-occupied houses (N = 1,286), the natural log of the median cash rent (N = 1,197), the percentage of the census block group’s estimated population who live in poverty (reverse-coded, N = 1,324), and the percentage of the census block group’s estimated population over the age of 24 who has less than a high school education⁴ (reverse-coded, N = 1,324). Figure 1 provides a visual depiction of the SES index variable across all census block groups in the analysis.

Figure 1.

The socioeconomic status index of Philadelphia, Pennsylvania, residents by census block group (N = 1,324). Data Source: U.S. Census Bureau.

Residential stability index

Less residential stability in an area often results in more reported crime (Bursik 1986, 1988; Skogan 1990). This index variable (N = 1,324), included to control for that effect, is the average of five standardized (z-scored) variables (Cronbach’s α = .874): the percentage of the census block group’s estimated population who lived in the same house in the previous year (N = 1,324), the percentage of the census block group’s estimated population who moved in before the year 2010 (N = 1,324), the percentage of the census block group’s estimated population who moved in before the year 2000 (N = 1,324), the median year that residents of the census block group moved in (reverse-coded, N = 1,322), and the percentage of the census block group’s estimated population who are renters (reverse-coded, N = 1,324).

Percentage White non-Hispanic

A person’s knowledge of their activity spaces informs their decisions as to where to commit crime (Brantingham and Brantingham 1993a). It has been shown that the size of one’s activity space is not consistent across races and ethnicities (Rengert and Groff 2011) and the capability of guardianship varies across areas of different racial compositions (Rengert et al. 2015). Furthermore, Pratt and Cullen (2005) conducted a meta-analysis which identified the socioeconomic predictors which were consistently the strongest in predicting crime; the percent non-White and the percent Black were separately identified as being strong predictors of crime, and the racial heterogeneity index was identified as being moderately strong across studies. In the interest of parsimony, this study used the percentage of each census block group’s estimated population who identify as White non-Hispanic to control for these effects (N = 1,324).

Estimated population

The estimated population is the total (estimated) population of the census block (N = 1,324).

The built environment—land use

The types of places within an area have impacts upon the counts, as well as the different types, of crimes that occur in those areas (Clarke and Eck 2007). The total number of parcels or units of land utilized for commercial, residential, and civic/institutional use in each census block group were included as three separate variables (N = 1,324 for all three of these variables).

Traffic and activity—street types

Routine activities often dictate where motivated offenders and suitable targets are likely to meet at the same place and time, thus permitting a crime to occur (Cohen and Felson 1979); this includes not only the destinations (nodes) where people are heading from and to but also the paths along which they travel, expanding their awareness and activity spaces (Brantingham and Brantingham 1993b). The total length, in miles, of four different types of streets within each census block group was included as separate variables in the analysis.⁵ It was necessary to differentiate between the different types of streets in the analysis, as each type is expected to carry different quantities of people upon them, at different speeds, and thus will permit varying levels of increased activity and awareness between them. These four types of street segments included in order of increasing frequency of traffic and activity: collector-residential streets, local residential streets, minor arterials, and major arterials (N = 1,324 for all four variables); street types without a strong theoretical nexus with reported crime, such as elevated interstate limited-access expressways, were excluded.

Intersections

The count of intersections regulated by a traffic light (or signal) were included as one variable in the analyses, and the count of intersections regulated by one or more stop signs were included as a separate variable. Stops at red lights are potentially much longer in duration than stops at stop signs, providing a person a far greater opportunity to expand their awareness space in the areas surrounding intersections regulated by traffic lights than those regulated only by stop signs; as such, the two may have different impacts upon the types and quantities of crime that occur in their respective areas.

Subway and elevated train stations

The locations of subway and elevated stations are certainly activity nodes (Haberman, Sorg, and Ratcliffe 2018; Sorg and Taylor 2011) but could arguably also be considered paths; regardless, the influx of persons carrying out their routine activities upon them required proper control in the analyses. As such, the count of subway and elevated train stations in each census block group was included (as one variable) in the analyses.

Spatial lag variables

A series of 10 global Moran’s I tests revealed that the counts of each of the 10 dependent variables demonstrated statistically significant (all p < .001) positive spatial autocorrelation, indicating that census block groups that are closer to each other are more related than are distant census block groups (Fotheringham 2009). Consistent with methods used to control for this concern in existing research, spatial lag variables were generated and included to account for the effect that crime in the immediate neighboring census block groups may have upon crime counts in the observed census block group (McCord and Ratcliffe 2007). These variables were constructed in GeoDa Version 1.8.14 using single-order, queen contiguity of the counts of each individual crime in neighboring census block groups to create 10 separate spatial lag variables, one for each of the 10 dependent crime variables.

High–low SES outliers (indicator variable)

The hypothesis postulated by this study required that census block groups that are outliers in terms of higher SES, as compared to their neighbors, be identified so that any effect in these areas may be captured in the analyses. As no such indicator for these areas is readily available, it was created using the spatial statistics tools available in ArcGIS Version 10.4 software. A local indicator of spatial association test, the local Moran’s I (Anselin 1995), was conducted of the SES index variable. A fixed distance of 10,560 feet (two miles) was selected, using Euclidean distance, so that the measure of SES of each observation (census block group) would be compared only to neighboring observations that fall within 10,560 feet of the observed census block group. The selection of this fixed distance band was grounded in both theory and pragmatism. Research is mixed upon the mean and median distances one may travel from home to commit crime, but they have been shown to vary according to several factors including age, geography, the type of crime, and “opportunity structure” (Ackerman and Rossmo 2015:242-43; Andresen, Frank, and Felson 2014). Generally, offenders have been shown to travel no more than a few miles to commit most crimes (Rossmo 2000; Townsley and Sidebottom 2010). Pragmatically, the ArcGIS Version 10.4 software indicated that a minimum distance threshold of 4,991 feet was necessary just to ensure that each and every census block group had at least one “neighbor” group for comparison, and ideally (for internal validity), a census block group should be compared to several neighboring census block groups to ensure only true outliers are identified and included in the analyses.

The local Moran’s I test identified 32 observations, of the 1,324 census block groups in the analysis, that were statistically significant high–low outliers in terms of SES compared only to neighboring census block groups within two miles of the census block group being observed. These outliers were marked and identified with a dichotomous dummy variable (SES-HL) for further analysis. The smallest of these 32 census block groups is 0.024 square miles (roughly equivalent to a square of approximately two city blocks × two city blocks⁶), the median is 0.035 square miles (slightly smaller than 2.5 blocks × 2.5 blocks), and the largest 0.088 square miles (slightly smaller than four blocks × four blocks). These outlier census block groups are much smaller than the typical census block groups included in the full analysis (SES-HL outliers have mean and median surface areas of 0.042 and 0.035 square miles, respectively, as opposed to 0.085 and 0.048 square miles for all census block groups), suggesting greater portions of many of these areas may be more consistent with the definitions of an edge discussed earlier. Figure 2 provides a visual depiction of the results of the local Moran’s I test. The locations of the census block groups identified as significant outliers and clusters of SES are shown.

Figure 2.

The Moran’s I local indicators of spatial association test compared the socioeconomic status (SES) index value of each observed census block group to neighboring census block groups within a two-mile band, which identified clusters and outliers of high and low SES. The inset map on the right details the area in which all 32 high–low outliers were identified. Data Source: U.S. Census Bureau.

Analysis

As discussed, a local Moran’s I test was conducted, which resulted in the SES-HL indicator variable; all data were then entered into Stata/IC Version 14.2 for full analysis. The 10 dependent variables were comprised of count data, and preliminary inspection revealed overdispersion in the distributions; as such, negative binomial regression was the preferred model for analysis (Rabe-Hesketh and Skrondal 2012). A total of 10 negative binomial regression analyses were conducted (one for each dependent variable). The same predictors were used in each of the 10 models, except as to the spatial lag variable, in which only the spatial lag variable that corresponded to that specific dependent crime variable was included. Models 1–5 examined rational crimes (in which an economic benefit exists to the offender), and models 6–10 examined nonrational crimes. A conservative Bonferroni-corrected p value threshold of .005 was used to control for the family-wise error rate (Dunn 1961).⁷

Results

Table 2 shows the full results of the negative binomial regression models for each dependent variable. Coefficients and standard errors are reported, followed by incident rate ratios (IRRs) which were also calculated for each predictor. Each of the models performed generally well overall. The Cragg and Uhler’s (pseudo) R ² values⁸ (Cragg and Uhler 1970) ranged from .49 to .66 for the first five models (the rational crimes) and from .36 to .64 for the second five models (the nonrational crimes). The performance of the 15 covariates varied to some extent by model (which is to be expected, given that causes and correlates of crime vary by the type of crime). The spatial lag variables proved to be especially consistent, strong predictors (p < .001 for all spatial lag variables); the IRRs for these variables were all greater than 1 (except for in the thefts from automobile model), indicating that the count of each crime is expected to increase as the count of the same crime increases in neighboring census block groups. The estimated population of each census block group was statistically significant (p < .005 or better) in each of the models except one (arson), with IRRs greater than 1, showing that counts of each crime (except arson) will increase as estimated population increases. The SES index, which is of particular interest to this study, was significant in every model except residential burglary (p < .005 or better), and each IRR was less than 1, indicating that as SES increases in these areas, the frequencies of almost all the crimes being studied decreases. The other covariates varied in strength and reliability across models, but in each model, at least four covariates were statistically significant predictors of the dependent crime variable.

Table 2.

Results of Negative Binomial Regression Models.

	Model 1			Model 2			Model 3			Model 4			Model 5
Rational Crimes Models	Robberies by Firearm			Robberies Other Than by Firearm			Residential Burglaries			Thefts from Automobiles			Automobile Thefts
Variable	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR
High–low SES outlier (“SES-HL”)	0.340*	.104	1.404	0.078	.113	1.081	0.085	.074	1.088	0.215	.079	1.240	0.236	.088	1.266
Lagged crime variable (specific to crime)	0.034**	.003	1.034	0.023**	.002	1.023	0.018**	.001	1.019	0.004**	.000	1.004	0.031**	.002	1.032
Population (each 100 people)	0.016**	.003	1.016	0.019**	.004	1.019	0.028**	.002	1.028	0.021**	.003	1.022	0.026**	.003	1.026
Percent White non-Hispanic	−0.747**	.069	0.474	−0.226*	.070	0.797	−0.205**	.046	0.815	0.430**	.051	1.538	−0.056	.056	0.945
SES index	−0.119**	.029	0.888	−0.180**	.031	0.835	−0.045	.020	0.956	−0.066*	.021	0.937	−0.134**	.025	0.875
Stability index	−0.068*	.021	0.934	−0.085**	.023	0.918	−0.075**	.015	0.928	−0.108**	.017	0.898	−0.016	.019	0.984
Civic/institutional land uses	0.016**	.004	1.016	0.028**	.004	1.028	0.004	.003	1.004	0.012**	.003	1.012	0.012**	.004	1.012
Commercial land uses (each 100 uses)	0.160	.070	1.174	0.184	.077	1.202	0.026	.049	1.026	0.078	.055	1.081	0.078	.060	1.081
Residential land uses (each 100 uses)	0.015	.009	1.015	0.014	.010	1.014	0.019*	.007	1.020	0.012	.007	1.012	0.016	.008	1.016
Intersections with traffic lights	0.024**	.003	1.025	0.028**	.004	1.028	.002	.002	1.002	0.018**	.002	1.018	0.015**	.003	1.015
Intersections with stop signs	0.008**	.002	1.008	0.006*	.002	1.006	0.005**	.001	1.005	0.005**	.001	1.005	0.006**	.001	1.006
Subway/elevated stations	0.173	.071	1.189	0.528**	.082	1.696	0.023	.051	1.023	0.065	.057	1.068	−0.011	.059	0.990
Major arterial miles	−0.182*	.054	0.834	−0.125	.057	0.882	−0.047	.037	0.954	−0.021	.041	0.979	−0.079	.044	0.924
Minor arterial miles	−0.098	.044	0.907	−0.065	.047	0.937	0.018	.032	1.018	0.059	.034	1.061	−0.001	.036	0.999
Local street miles	−0.144**	.038	0.866	−0.154**	.040	0.857	0.018	.025	1.019	0.000	.027	1.000	−0.059	.030	0.942
Collector street miles	0.094	.037	1.098	0.125*	.039	1.133	0.058	.026	1.060	0.236**	.028	1.266	0.155**	.030	1.167
Constant	1.513**	.068	—	1.558**	.066	—	2.243**	.051	—	2.522**	.045	—	1.316**	.052	—
Ln(α)	−1.528	.059	—	−1.255	.049	—	−2.055	.049	—	−1.772	.043	—	−1.831	.057	—
Cragg and Uhler’s R ²	.55			.56			.49			.66			.56
Nonrational Crimes Models	Model 6			Model 7			Model 8			Model 9			Model 10
Nonrational Crimes Models	Aggravated Assaults by Firearm			Aggravated Assaults Other Than by Firearm			Vandalisms			Arsons			Rapes
Variable	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR	Coef.	SE	IRR
High–low SES outlier (“SES-HL”)	0.181	.139	1.199	0.171	.098	1.186	0.094	.064	1.098	−0.342	.250	0.711	0.043	.152	1.043
Lagged crime variable (specific to crime)	0.042**	.004	1.043	0.017**	.002	1.017	0.007**	.000	1.007	0.127**	.017	1.135	0.076**	.009	1.079
Population (each 100 people)	0.020**	.004	1.020	0.020**	.003	1.020	0.024**	.002	1.025	0.011	.005	1.011	0.031**	.004	1.031
Percent White non-Hispanic	−1.395**	.086	0.248	−0.424**	.059	0.654	−0.073	.039	0.930	−0.539**	.119	0.583	−0.778**	.087	0.459
SES index	−0.323**	.036	0.724	−0.300**	.029	0.741	−0.127**	.018	0.881	−0.372**	.050	0.689	−0.275**	.037	0.760
Stability index	0.025	.026	1.025	−0.003	.019	0.997	−0.003	.013	0.997	−0.073	.038	0.930	−0.053	.027	0.948
Civic/institutional land uses	0.009	.005	1.009	0.015**	.004	1.016	0.011**	.003	1.011	0.005	.007	1.005	0.017**	.005	1.018
Commercial land uses (each 100 uses)	0.190	.084	1.209	0.130	.065	1.139	0.076	.044	1.079	0.291	.104	1.338	0.035	.084	1.036
Residential land uses (each 100 uses)	−0.004	.011	0.996	0.009	.009	1.009	0.009	.006	1.009	0.022	.015	1.022	0.005	.011	1.005
Intersections with traffic lights	0.021**	.004	1.021	0.021**	.003	1.021	0.014**	.002	1.015	0.011	.005	1.011	0.016**	.004	1.016
Intersections with stop signs	0.009**	.002	1.009	0.004	.002	1.004	0.006**	.001	1.006	0.004	.002	1.004	0.004	.002	1.004
Subway/elevated stations	0.122	.079	1.129	0.203*	.066	1.225	0.054	.043	1.055	−0.028	.105	0.972	0.217	.077	1.243
Major arterial miles	−0.132	.063	0.877	−0.108	.048	0.898	−0.116**	.031	0.890	0.039	.085	1.040	−0.052	.064	0.950
Minor arterial miles	−0.108	.051	0.898	−0.037	.039	0.964	0.005	.026	1.005	0.101	.065	1.106	−0.024	.049	0.977
Local street miles	−0.027	.044	0.973	−0.022	.033	0.979	−0.023	.022	0.978	0.149	.055	1.161	−0.069	.046	0.933
Collector street miles	−0.018	.045	0.983	0.098*	.034	1.103	.081**	.022	1.084	0.234**	.055	1.264	0.032	.045	1.033
Constant	1.284**	.076	—	2.009**	.063	—	2.941**	.043	—	−0.604**	.098	—	0.435**	.076	—
Ln(α)	−1.350	.066	—	−1.548	.051	—	−2.283	.046	—	−1.261	.124	—	−1.764	.105	—
Cragg and Uhler’s R ²	.64			.57			.57			.36			.48

Note: N = 1,324. SES = socioeconomic status; Coef. = coefficient; SE = standard error; IRR = incident rate ratio.

*p < .005. **p < .001.

The results of the SES-HL indicator variable (the “socioeconomic edges”) in the models showed that it was a statistically significant predictor of robberies by firearm (IRR = 1.404, p < .005), but not in any of the other four models for rational crimes. These results indicate that census block groups shown to be statistically significant outliers of high SES, situated in an area of predominantly lower SES, can expect to experience robberies by firearm at a 40 percent greater rate than would otherwise be expected, holding all over variables constant. As hypothesized, the areas of higher SES located within or near areas of predominantly lower SES did not experience greater nonrational crime.

Discussion

The model which predicted robberies by firearm, a rational crime in which there exists an economic incentive, demonstrated that at a socioeconomic edge of higher SES that rests in a much larger area of comparatively lower SES, 40 percent more of these crimes occurred than would otherwise be expected. This particular result was hypothesized, consistent with the theoretical framework discussed earlier. Offenders seek the greatest rewards and the lowest risk (Clarke and Cornish 1985); if these areas are in an offender’s awareness space (Brantingham and Brantingham 1993b), they will recognize the potentially higher economic rewards, more easily assess the risks, and assign appropriately higher opportunity values to the areas of the socioeconomic edges. The hypothesis was further supported in that the socioeconomic edges did not experience increases of any nonrational crimes, which were included as nonequivalent dependent variables to demonstrate that any observed criminogenic effect in the socioeconomic edges is specific only to economically driven rational crimes. Although each of the models predicting nonrational crimes had generally performed well, and the overall level of SES in the census block group was a significant predictor in each of the nonrational crime models (decreases in SES resulted in increases in each type of nonrational crime), the socioeconomic edges did not experience any greater nonrational crime than would otherwise be expected. This result was hypothesized as the edges offered enhanced economic rewards but do not offer enhanced rewards to offenders of rape, arson, vandalism, or any type of aggravated assault.

A significant socioeconomic edge effect was observed for one rational crime; however, none of the three rational property crimes increased in frequency as hypothesized,⁹ despite the economic rewards being potentially greater in these areas, and there are a few possible explanations as to why. It is possible that the areas may not be in many offenders’ awareness or activity spaces, making them unaware of the potentially increased rewards; this, however, is unlikely given the short fixed-distance band used in the study. Another possible explanation for these findings may be that the fixed distance band used in this study is more consistent with the criminal journeys of violent (gunpoint robbers) rather than property offenders. If true, this would then be consistent with the research of Hesseling (1992), who found that violent offenders are more likely than property offenders to commit their crime in their own neighborhood and also indicated that increased violent crime “in lower-SES neighborhoods are partly due to the residence of more offenders in these neighborhoods” (p. 109). The distance an offender will travel to commit crime has been shown to vary by type of crime (Ackerman and Rossmo 2015; Andresen et al. 2014; Rossmo 2000; Townsley and Sidebottom 2010). The fixed distance band used in this study’s local Moran’s I test to identify high–low SES outliers was based upon findings in journey-to-crime research, but it did not vary by crime type.

A more probable explanation then for these findings rests not in the reward but instead in the risk side of the equation, as the potentially enhanced economic rewards for the commission of these property crimes may not sufficiently outweigh what may also be perceived as a potentially increased risk in these areas. Cornish and Clarke (1987) have indicated that offenders balance a number of factors in their decision-making, and although they identify the reward value, “likely cash yield per crime,” as one factor to be considered, they also identify numerous risk-side factors as well including “risk of apprehension,” “severity of punishment,” “physical danger,” “instrumental violence required,” and “confrontation with victim” (p. 940, table 1). The socioeconomic edges represent an enhanced reward value in rational crimes, but the value of these rewards, and the types and costs of any associated risks, vary considerably by crime type, a variance likely only to be amplified when viewed by an offender in light of the many other environmental and situational considerations in their decision-making “template” as outlined in crime pattern theory (Brantingham and Brantingham 1993a:264). It was expected that the effect of the socioeconomic edge would then vary by crime type, as this was precisely why crimes were examined in this study as separate dependent variables in separate models.

One considerable risk-side factor that may potentially be of greater relevance to property crime than violent crime is capability of guardianship,¹⁰ and how that factors into offenders’ perceptions of overall risk in terms of being identified and/or arrested. Greater economic standing comes with greater ability to have things that help protect one’s property when one is away, such as security or doorbell cameras, gates, and alarms; these tools target-harden and increase guardianship, basic tenets of situational crime prevention (Cornish and Clarke 2003). This contention would be consistent with the research of Rengert and Groff (2011) and Rengert and Wasilcheck (1985) who found that residential burglars prefer to operate in areas wealthier than their own, but some burglars consider the criminal risk in some wealthier areas to be so great as to preclude them from consideration altogether; the same may hold true of other rational property crimes as well.

The null results in the robberies by other than firearm model may be the most unexpected, particularly in light of the strong significant effect in the robberies by firearm model. The decision to commit a robbery with any weapon (firearm or any other) requires at least three separate decision points; first, one must make a criminal involvement decision to commit a robbery; second, one must decide to arm oneself with a weapon in order to carry out the crime; and finally, one must make a criminal event decision. The Brantinghams (1993a) have indicated that the “number and sequence of decision points” in criminal decision-making is not consistent across crime types (p. 262), and Clarke and Cornish (1985) have stated that the decision-making process to commit some crimes will be “lengthier” than others (p. 169). The additional preparation step that requires arming oneself may imply that more forethought and planning be made prior to the commission of any robbery involving a weapon than for robberies in which no weapon was used and which may occur at any time opportunity strikes. This additional forethought may include a more comprehensive risk versus reward analysis, leading one toward the areas of greater economic means in their awareness space. The commission of a robbery using a firearm carries with it enhanced criminal penalties (and thus, risk) as well as greater physical risk. This may necessitate in one’s mind that the rewards must be much greater to compensate for the increased risk, driving the gunpoint robbers to these specific areas but not the nongunpoint robbers who may not perceive the additional risk and therefore may not necessarily seek the greater reward to compensate for it.

Limitations

There exist methodological limitations in this study, derived primarily from the data utilized. First, there are concerns that are inherent with utilizing official police data, foremost of which is the concern that some crimes are not reported to police (Gottfredson and Gottfredson 1988). Although a five-year cross section of police data is not going to capture every crime incident that occurred, it has been demonstrated that the spatial patterns of the missing data will typically be highly correlated with the spatial patterns of the reported incidents of the same crimes (Chainey and Ratcliffe 2005); thus, this is not cause for great concern. Another limitation is that of the modifiable areal unit problem (for a detailed explanation, see Chainey and Ratcliffe 2005). Although this study is conducted at a small unit of analysis to achieve the greatest possible homogeneity within the units of analysis in terms of population, SES, residential stability, and the other covariates, heterogeneity will still exist within many of the census block groups.

Another limitation rests in the use of ACS data, as high error margins relative to estimated values do exist in some census block groups for some socioeconomic and demographic data. Additionally, a visual inspection shows that some of the areas identified as higher SES outliers have experienced various degrees of recent gentrification. ACS data are available at the census block group level only as five-year estimates and fail to capture shorter term longitudinal changes that may have occurred between 2010 and 2014. Although the extant research into crime and gentrification indicates that the relationship between the two, if one exists, is unclear as findings vary,¹¹ and there is presently no scientific accord as to the precise definition of gentrification, what constitutes a gentrified neighborhood, or how to measure or control for it (Barton 2016), it cannot be entirely ruled out that it may have had some impact upon the results.

Finally, it could be argued that a limitation exists in the selection of the SES-HL outliers as a dichotomous indicator variable, as one might contend that the significant differences in z-scores between an observation and the more general area in which it is located that are marked as outliers by the local Moran’s I test might better be reflected as continuous variables (i.e., perhaps using standard residuals). Although on the surface, this appears to be a valid methodological concern, the use of a dichotomous variable identifying only the outliers is preferable as it ensures that only those census block groups which stand in the starkest contrast to their surrounding areas, which are thus those that are also the most likely to be perceived as such to potential offenders when assessing potential rewards in their criminal event decisions, are included in the analysis. Although smaller differences in SES between census block groups do represent potentially different rewards, they may not be as readily perceived as apparent outliers, as existing research has argued that a person may not initially distinguish “subtle” changes from one area to a next, even if their journey takes them through or over an edge (Song et al. 2017:2).

Implications for Theory

Despite the overall mixed results, there is a substantive finding in that one form of rational crime did experience a considerable increase in frequency in the socioeconomic edges. In just the past several years, Song et al. (2017) found criminogenic edges at changes in land use, Rengert et al. (2015) at the edges of communities of different races, and Song et al. (2013) at the edges of residential zoning changes. This study builds upon the literature on edges in crime pattern theory by arguing that there exists a criminogenic effect at edges of SES as well, in that small edges of higher SES situated in larger areas of predominantly and comparatively lower SES did experience increased robberies by firearm. This effect was driven by offenders who viewed the enhanced rewards present in these edges, through the decision-making template outlined in crime pattern theory that simultaneously considers risks, rewards, and situational factors, and determined that the enhanced rewards in these edge areas represented enhanced criminal opportunities for robberies by firearm. Offenders of other rational crimes did not find that the enhanced rewards of the edges represented better criminal opportunities, suggesting other situational or risk-side factors specific to the other types of rational crimes may have outweighed the benefits of the edge.

It has been well established by research in social disorganization theory (Shaw and McKay 1942) that access to economic resources is generally correlated with lower crime, as communities and neighborhoods with higher SES typically have a greater ability to exercise informal social control (Bursik 1988; Kornhauser 1978), which is often activated through collective efficacy (Sampson 2012). Had it not been for Shaw and McKay’s groundbreaking work in social disorganization theory, opportunity theories and environmental criminology may not even exist; despite this, much remains to be learned about how these theories interact with one other. In the present study, the small edge areas have higher SES but are located in larger areas that predominantly and comparatively have much lower SES, and it is unknown how much influence the high SES of the edge areas, or the low SES of the surrounding neighborhood, had upon individuals’ risk assessments. The high SES may have deterred some offenders, such as those for property crimes, while gunpoint robbers may have perceived the low SES of the wider area and been less dissuaded. This raises the possibility of an interaction between individual decision-making and spatial crime patterns, as well as possible moderating effects at a neighborhood level, at what would amount to a three-way intersection of crime pattern, rational choice, and social disorganization theories. The relationships and interactions between theories at different levels of analysis are complicated and often difficult to study using available data (Andresen and Malleson 2011; Birks, Townsley, and Stewart 2012; Coleman 1990; Sampson 2012, Sampson and Groves 1989). Crime pattern theory does attempt to bridge some of this analytical gap (Johnson 2010), but as other recent research has either found and/or suggested, there is still a great deal that we may learn at the junction of these theories that traditionally operate or are studied at different levels of analysis (Braga and Clarke 2014; Nobles, Ward, and Tillyer 2016; O’Brien 2019; Sullivan and McGloin 2014; Weisburd, Groff, and Yang 2012).

The findings of this study open other questions for future research. The study finds a “socioeconomic edge” effect that exists in small outlier areas of SES, but the questions as to whether such a strong effect might persist between areas that are not in such sharp contrast in terms of SES, or whether similar effects may exist at the points of distinction of other types of socioeconomic variables, remain unanswered. Additionally, it is unknown how potentially moderating variables at different levels of analysis, such as collective efficacy, function in these socioeconomic edges and what impact this may have upon different types of reported crime.

Footnotes

Acknowledgments

The author wishes to thank Professors Jerry Ratcliffe, Elizabeth Groff, and Ralph Taylor, Co-Editor Christopher J. Sullivan, and the anonymous reviewers for all their helpful guidance and comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Robert Drew Heinzeroth, Sr.

Notes

References

Ackerman

Jeffrey M.

Kim Rossmo

. 2015. “How Far to Travel? A Multilevel Analysis of the Residence-to-crime Distance.” Journal of Quantitative Criminology 31:237–62.

Andresen

Martin A.

Frank

Richard

Felson

Marcus

. 2014. “Age and the Distance to Crime.” Criminology & Criminal Justice 14:314–33.

Andresen

Martin A.

Malleson

Nicolas

. 2011. “Testing the Stability of Crime Patterns: Implications for Theory and Practice.” Journal of Research in Crime and Delinquency 48:58–82.

Anselin

Luc

. 1995. “Local Indicators of Spatial Association—LISA.” Geographic Analysis 27:93–115.

Barton

Michael S.

2016. “Gentrification and Violent Crime in New York City.” Crime & Delinquency 62:1180–202.

Birks

Daniel

Townsley

Michael

Stewart

Anna

. 2012. “Generative Explanations of Crime: Using Simulation to Test Criminological Theory.” Criminology 50:221–54.

Braga

Anthony A.

Clarke

Ronald V.

. 2014. “Explaining High-risk Concentrations of Crime in the City: Social Disorganization, Crime Opportunities, and Important Next Steps.” Journal of Research in Crime and Delinquency 51:480–98.

Brantingham

Patricia L.

Brantingham

Paul J.

. 1975. “Residential Burglary and Urban Form.” Urban Studies 12:273–84.

Brantingham

Patricia L.

Brantingham

Paul J.

. 1993a. “Environment, Routine, and Situation: Toward a Pattern Theory of Crime.” Pp. 259–94 in Routine Activity and Rational Choice, Advances in Criminological Theory, edited by Clarke

Ronald V.

Felson

Marcus

. New Brunswick, NJ: Transaction.

10.

Brantingham

Patricia L.

Brantingham

Paul J.

. 1993b. “Nodes, Paths and Edges: Considerations on the Complexity of Crime and the Physical Environment.” Journal of Environmental Psychology 13:3–28.

11.

Brantingham

Patricia L.

Brantingham

Paul J.

. 1995. “Criminality of Place: Crime Generators and Crime Attractors.” European Journal on Criminal Policy and Research 3:5–26.

12.

Brantingham

Patricia L.

Brantingham

Paul J.

. 1997. “Mapping Crime for Analytic Purposes: Location Quotients, Counts and Rates.” Pp. 263–88 in Crime Mapping and Crime Prevention, edited by Weisburd

David

McEwen

Tom

. Monsey, NY: Criminal Justice Press.

13.

Brantingham

Paul J.

Brantingham

Patricia L.

. 1991. “Introduction to the 1991 Reissue: Notes on Environmental Criminology.” Pp. 1–6 in Environmental Criminology, edited by Brantingham

Paul J.

Brantingham

Patricia L.

. Prospect Heights, IL: Waveland Press.

14.

Brantingham

Paul J.

Dyreson

Delmar A.

Brantingham

Patricia L.

. 1976. “Crime Seen Through a Cone of Resolution.” American Behavioral Scientist 20:261–73.

15.

Braveman

Paula A.

Cubbin

Catherine

Egerter

Susan

Chideya

Sekai

Marchi

Kristen S.

Metzler

Marilyn

Posner

Samuel

. 2005. “Socioeconomic Status in Health Research: One Size Does Not Fit All.” JAMA 294:2879–88.

16.

Bursik

Robert J.

1986. “Ecological Stability and the Dynamics of Delinquency.” Pp. 35–66 in Crime and Justice, Volume 8: Communities and Crime, edited by Reiss

Albert J.

Jr Tonry

Michael

. Chicago: University of Chicago Press.

17.

Bursik

Robert J.

1988. “Social Disorganization and Theories of Crime and Delinquency: Problems and Prospects.” Criminology 26:519–51.

18.

Campbell

Colin

. 2015. “The Socioeconomic Consequences of Dropping Out of High School: Evidence from an Analysis of Siblings.” Social Science Research 51:108–18.

19.

Chainey

Spencer

Ratcliffe

Jerry H.

. 2005. GIS and Crime Mapping. London, England: Wiley and Sons.

20.

Clarke

Ronald V.

Cornish

Derek B.

. 1985. “Modeling Offender’s Decisions: A Framework for Research and Policy.” Crime and Justice 6:147–85.

21.

Clarke

Ronald V.

Eck

John E.

. 2007. Understanding Risky Facilities. Problem-oriented Guides for Police, Problem-solving Tools Series, Guide No. 6. Washington, DC: U.S. Department of Justice.

22.

Cohen

Lawrence E.

Felson

Marcus

. 1979. “Social Change and Crime Rate Trends: A Routine Activity Approach.” American Sociological Review 44:588–608.

23.

Coleman

James S.

1990. Foundations of Social Theory. Cambridge, MA: Harvard University Press.

24.

Cornish

Derek B.

Clarke

Ronald V.

. 1986. “Introduction.” Pp. 1–16 in The Reasoning Criminal: Rational Choice Perspectives on Offending, edited by Cornish

Derek B.

Clarke

Ronald V.

. New York: Springer-Verlag.

25.

Cornish

Derek B.

Clarke

Ronald V.

. 1987. “Understanding Crime Displacement: An Application of Rational Choice Theory.” Criminology 25:933–47.

26.

Cornish

Derek B.

Clarke

Ronald V.

. 2003. “Opportunities, Precipitators, and Criminal Decisions: A Reply to Wortley’s Critique of Situational Crime Prevention.” Pp. 41–96 in Theory for Practice in Situational Crime Prevention, edited by Smith

Martha J.

Cornish

Derek B.

. Monsey, NY: Criminal Justice Press.

27.

Covington

Jeanette C.

Taylor

Ralph B.

. 1989. “Gentrification and Crime: Robbery and Larceny Changes in Appreciating Baltimore Neighborhoods During the 1970s.” Urban Affairs Quarterly 25:142–72.

28.

Cragg

John G.

Uhler

Russell S.

. 1970. “The Demand for Automobiles.” The Canadian Journal of Economics 3:386–406.

29.

DeMaris

Alfred

. 2002. “Explained Variance in Logistic Regression: A Monte Carlo Study of Proposed Measures.” Sociological Methods & Research 31:27–74.

30.

Dunn

Olive J.

1961. “Multiple Comparisons among Means.” Journal of the American Statistical Association 56:52–64.

31.

Fotheringham

A. Stewart

. 2009. “‘The Problem of Spatial Autocorrelation’ and Local Spatial Statistics.” Geographical Analysis 41:398–403.

32.

Gottfredson

Michael R.

Gottfredson

Don M.

. 1988. Decision Making in Criminal Justice: Toward the Rational Exercise of Discretion. 2nd ed. New York: Plenum Press.

33.

Groff

Elizabeth R.

Lockwood

Brian

. 2014. “Criminogenic Facilities and Crime across Street Segments in Philadelphia: Uncovering Evidence about the Spatial Extent of Facility Influence.” Journal of Research in Crime and Delinquency 51:277–314.

34.

Guerry

Andre-Michel

. 1833. Essai sur la Statistique Morale de la France: Precede d’un Rapport a l’Academie de Sciences. Paris, France: Chez Crochard.

35.

Haberman

Cory P.

Sorg

Evan T.

Ratcliffe

Jerry H.

. 2018. “The Seasons They Are a Changin’: Testing for Seasonal Effects of Potentially Criminogenic Places on Street Robbery.” Journal of Research in Crime and Delinquency 55:425–59.

36.

Hesseling

Rene B. P.

1992. “Using Data on Offender Mobility in Ecological Research.” Journal of Quantitative Criminology 8:95–112.

37.

Johnson

Shane D.

2010. “A Brief History of the Analysis of Crime Concentration.” European Journal of Applied Mathematics 21:349–70.

38.

Kornhauser

Ruth R.

1978. Social Sources of Delinquency. Chicago: University of Chicago Press.

39.

Lersch

Kim M.

2007. Space, Time, and Crime. 2nd ed. Durham, NC: Carolina Academic Press.

40.

McCord

Eric S.

Ratcliffe

Jerry H.

, 2007. “A Micro-spatial Analysis of the Demographic and Criminogenic Environment of Drug Markets in Philadelphia.” The Australian and New Zealand Journal of Criminology 40:43–63.

41.

McDonald

Scott C.

1986. “Does Gentrification Affect Crime Rates?” Crime and Justice 8:163–201.

42.

Nobles

Matt R.

Ward

Jeffrey T.

Tillyer

Rob

. 2016. “The Impact of Neighborhood Context on Spatiotemporal Patterns of Burglary.” Journal of Research in Crime and Delinquency 53:711–40.

43.

O’Brien

Daniel T.

2019. “The Action Is Everywhere, but Greater at More Localized Spatial Scales: Comparing Concentrations of Crime across Addresses, Streets, and Neighborhoods.” Journal of Research in Crime and Delinquency 56:339–77.

44.

Papachristos

Andrew V.

Smith

Chris M.

Scherer

Mary L.

Fugiero

Melissa A.

. 2011. “More Coffee, Less Crime? The Relationship between Gentrification and Neighborhood Crime Rates in Chicago, 1991 to 2005.” City & Community 10:215–40.

45.

Perneger

Thomas V.

1998. “What’s Wrong with Bonferroni Adjustments.” BMJ 316:1236–38.

46.

Pratt

Travis C.

Cullen

Francis T.

. 2005. “Assessing Macro-level Predictors and Theories of Crime: A Meta-analysis.” Crime and Justice 32:373–450.

47.

Rabe-Hesketh

Sophia

Skrondal

Anders

. 2012. Multilevel and Longitudinal Modeling Using Stata. 3rd ed. College Station, TX: Stata Press.

48.

Rengert

George F.

Groff

Elizabeth R.

. 2011. Residential Burglary: How the Urban Environment and Our Lifestyles Play a Contributing Role. 3rd ed. Springfield, IL: Charles C. Thomas.

49.

Rengert

George F.

Lockwood

Brian

Groff

Elizabeth R.

. 2015. “Burglary in a Segregated City: Race of Offenders and Community of Offending.” Pp. 214–39 in The Criminal Act: The Role and Influence of Routine Activity Theory, edited by Andresen

Martin

Farrell

Graham

. New York: Palgrave Macmillian.

50.

Rengert

George F.

Wasilchick

John

. 1985. Suburban Burglary: A Time and a Place for Everything. Springfield, IL: Charles C. Thomas.

51.

Robinson

William S.

1950. “Ecological Correlations and the Behavior of Individuals.” American Sociological Review 15:351–57.

52.

Rossmo

D. Kim

. 2000. Geographic Profiling. Boca Raton, FL: CRC Press.

53.

Rothman

Kenneth J.

1990. “No Adjustments Are Needed for Multiple Comparisons.” Epidemiology 1:43–46.

54.

Sampson

Robert J.

2012. Great American City: Chicago and the Enduring Neighborhood Effect. Chicago: University of Chicago Press.

55.

Sampson

Robert J.

Byron Groves

. 1989. “Community Structure and Crime: Testing Social-disorganization Theory.” American Journal of Sociology 94:774–802.

56.

Shadish

William R.

Cook

Thomas D.

. 2009. “The Renaissance of Field Experimentations in Evaluating Interventions.” Annual Review of Psychology 60:607–29.

57.

Shadish

William R.

Cook

Thomas D.

Campbell

Donald T.

. 2002. Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston, MA: Houghton-Mifflin.

58.

Shaw

Clifford R.

McKay

Henry D.

. 1942. Juvenile Delinquency and Urban Areas. Chicago: University of Chicago Press.

59.

Sherman

Lawrence W.

Gartin

Patrick R.

Buerger

Michael E.

. 1989. “Hot Spots of Predatory Crime: Routine Activities and the Criminology of Place.” Criminology 27:27–55.

60.

Skogan

Wesley G.

1990. Disorder and Decline: Crime and the Spiral of Decay in America. New York: Free Press.

61.

Song

Justin

Andresen

Martin A.

Brantingham

Patricia L.

Spicer

Valerie

. 2017. “Crime on the Edges: Patterns of Crime and Land Use Change.” Cartography and Geographic Information Science 44:51–61.

62.

Song

Justin

Spicer

Valerie

Brantingham

Patricia L.

. 2013. “The Edge Effect: Exploring High Crime Zones near Residential Neighborhoods.” Paper presented at the 2013 IEEE International Conference on Intelligence and Security Informatics, Seattle, WA, USA, June 4–7, 2013.

63.

Sorg

Evan T.

Taylor

Ralph B.

. 2011. “Community Level Impacts of Temperature on Urban Street Robbery.” Journal of Criminal Justice 39:463–70.

64.

Sullivan

Christopher J.

McGloin

Jean Marie

. 2014. “Looking Back to Move Forward: Some Thoughts on Measuring Crime and Delinquency over the Past 50 Years.” Journal of Research in Crime and Delinquency 51:445–66.

65.

Taylor

Ralph B.

1988. Human Territorial Functioning: An Empirical, Evolutionary Perspective on Individual and Small Group Territorial Cognitions, Behaviors, and Consequences. New York: Cambridge University Press.

66.

Taylor

Ralph B.

1997. “Social Order and Disorder of Street Blocks and Neighborhoods: Ecology, Microecology, and the Systemic Model of Social Disorganization.” Journal of Research in Crime and Delinquency 34:113–55.

67.

Taylor

Ralph B.

Covington

Jeanette

. 1988. “Neighborhood Changes in Ecology and Violence.” Criminology 26:553–90.

68.

Townsley

Michael

Sidebottom

Aiden

. 2010. “All Offenders Are Equal, but Some Are More Equal Than Others: Variations in Journeys to Crime between Offenders.” Criminology 48:897–917.

69.

U.S. Census Bureau. 2015. American Community Survey. Washington, DC: U.S. Census Bureau.

70.

Weisburd

David L.

Eck

John E.

Braga

Anthony A.

Telep

Cody W.

Cave

Breanne

Bowers

Kate

Bruinsma

Gerben

Gill

Charlotte

Groff

Elizabeth R.

Hibdon

Julie

Hinkle

Joshua C.

Johnson

Shane D.

Lawton

Brian

Lum

Cynthia

Ratcliffe

Jerry H.

Rengert

George

Taniguchi

Travis

Yang

Sue-Ming

. 2016. Place Matters: Criminology for the Twenty-first Century. New York: Cambridge University Press.

71.

Weisburd

David L.

Groff

Elizabeth R.

Yang

Sue-Ming

. 2012. The Criminology of Place: Street Segments and Our Understanding of the Crime Problem. New York: Oxford University Press.

72.

Winters-Miner

Linda

Bolding

Pat S.

Hilbe

Joseph M.

Goldstein

Mitchell

Hill

Thomas

Nisbet

Robert

Walton

Nephi

Miner

Gary

. 2015. Practical Predictive Analytics and Decisioning Systems for Medicine: Informatics Accuracy and Cost-effectiveness for Healthcare Administration and Delivery Including Medical Research. London, England: Academic Press.